CES 2026 spotlights on Jensen Huang and NVIDIA, with new architectures Rubin and Alpamayo models announcing AI’s transition from screens to factories, streets, and laboratories.
(Background recap: Jensen Huang’s response in Taiwan to Google TPU competition: NVIDIA GPUs are more versatile, Nvidia is not afraid of market lag)
(Additional context: China’s version of Nvidia “Moore Thread” surged 468% on its debut day, early investors earning 6,200 times, creating a new A-share myth)

Table of Contents

• Rubin Platform: Cloud-based reasoning at the core
• Alpamayo: Enabling autonomous systems to start thinking
• From protein folding to climate simulation: a computing power frontier
• Policy trends and capital enthusiasm: a convergence point

Amid the chaos following the Trump administration’s invasion of Venezuela, the CES venue in Las Vegas remains crowded with people hopeful about technology. All eyes finally settle on a signature black leather jacket—NVIDIA CEO Jensen Huang. In his keynote, he straightforwardly states that the “screen era” of generative AI has peaked, and AI will soon “grow limbs and eyes,” officially entering the physical world.

Rubin Platform: Cloud-based reasoning at the core

To support real-time decision-making in ever-changing real-world scenarios, NVIDIA unveils the new Rubin platform. Rubin GPUs feature advanced packaging and are equipped with 288GB HBM4, with bandwidth reaching 22 TB/s; when 72 Rubin GPUs and 36 Vera CPUs are installed in the Vera Rubin NVL72 chassis, inference computing power can soar to 3,600 PFLOPS. According to NVIDIA official statements, mass production timelines are not yet fixed, but the technological threshold has already marked a deep divide in the market. Huang further emphasizes the concept of “scaling during testing,” indicating that training costs may stabilize, while inference demands will grow exponentially.

“This is not just an upgrade of chips; it’s the embodiment of AI.”

This single statement shifts the focus from silicon chips to streets and production lines, also signaling the next round of capital competition on Wall Street.

Alpamayo: Enabling autonomous systems to start thinking

Beyond hardware, NVIDIA’s open-source model family Alpamayo also becomes a key discussion point. The flagship version Alpamayo 1 (R1) brings “reasoning chain” capabilities into visual, language, and action integration models. When sudden road hazards like cracked guardrails or fallen cargo appear, R1 no longer just reacts reflexively but first infers the cause and then chooses the best avoidance route. Mercedes-Benz’s upcoming CLA model will be the first to adopt this system, with Uber and JLR as partners, marking a milestone where autonomous driving enters the “reasoning race” stage.

From protein folding to climate simulation: a computing power frontier

Huang Huang does not limit his scope to roads; he also cites examples like protein synthesis model OpenFold3 and Earth-2 climate platform. He has transformed DGX Cloud into a modern scientific “microscope.” Drug development and precise weather forecasting both require large-scale matrix computations, and the energy provided by the Rubin architecture allows these fields to experiment in real-time. In other words, NVIDIA is shifting from primarily selling graphics cards to becoming a provider of fundamental infrastructure for global scientific research and industrial automation.

Policy trends and capital enthusiasm: a convergence point

Against the backdrop of Trump’s emphasis on manufacturing return policies, Huang Huang champions “AI factories” and robotics solutions, successfully positioning himself in the domestic automation market. The repeated mention of “scaling during testing” in his speech paints a picture of an increasing demand curve for computing power. Market interpretation suggests that once Rubin chips enter mass production and Alpamayo models move out of labs, Industry 4.0 scenarios will turn from conceptual diagrams into mass-produced assembly lines. For capital markets, this computing expansion war is far from over.

The stage lights of CES 2026 gradually dim, but no one doubts NVIDIA’s ambition to let AI leave screens and illuminate more scenes in streets, factories, and labs. Next time in Las Vegas, perhaps rows of robots with silicon brains will greet visitors, and Wall Street will use an updated valuation model to price PFLOPS.